Malignant astrocytomas are among the most common and deadly brain tumors of childhood. Most affected children die within several years of diagnosis, despite current treatments; however, 20 to 30% respond favorably to therapy and are cured. The basis for these diverse outcomes has been enigmatic, even taking into account clinical and histological factors. In preliminary studies with an institutional cohort of pediatric gliomas, we observed that molecular markers could supplement histological data to refine prognostic assessments. Based on these findings, we began a more extensive study of the cohort of Children's Cancer Group study CCG-945, the largest group of pediatric high-grade gliomas accrued to date, and subsequently the Children's Oncology Group ACNS0126 study. During the previous funding period of this project, archival tumor specimens were obtained on more than 230 patients, the vast majority of which were evaluable in histological and genotyping analyses, the largest such analyses undertaken to date. The large size of these cohorts and the consistent treatment approaches used, coupled with the availability of central neuropathology review and comprehensive clinical data, provided a unique opportunity to address issues of molecular etiology and prognostic factors. Our studies demonstrated a striking association between outcome and several molecular features, including MGMT expression status, independent of clinical or histological factors; identified significant differences between molecular features of childhood and adult gliomas; and generated a sizeable resource of tumor tissue for further analyses. The proposed studies will use this resource as well as newly acquired tumor specimens and paired normal tissue, derived from 200 children treated on two new Children's Oncology Group high-grade glioma studies, to define the genetic alterations that characterize pediatric malignant gliomas, as a basis for prioritizing novel targets for prognostic and therapeutic stratification. We hypothesize that categorization of these tumors by their genomic alterations and drug resistance phenotype will improve accuracy of diagnostic and prognostic assessments, and provide insights into novel therapeutic targets. To test these hypotheses, we propose studies with the following aims: 1) Assess the frequency, composition, and prognostic relevance of alterations in Akt and MAPK pathway activation in pediatric malignant gliomas; 2) Determine whether O6-methylguanine- DNA methyltransferase (MGMT) overexpression and promoter methylation are associated with progression- free survival, independent of alkylator therapy; 3) Define, on a genome-wide basis, the spectrum and prognostic relevance of genomic alterations in pediatric malignant gliomas, using high-density SNP microarray- based profiling. Relevant markers will be evaluated in the context of conventional prognostic factors, such as histology, to determine their utility for biologically classifying childhood malignant gliomas.